THE NEXT THING, IN PERSPECTIVEIndependent AI news

The latest AI news, in focus.

Official announcements and industry reporting. Follow the timeline of AI.

Today's news

News feed
AI NEWS × FOXAIFind your next step

From the latest news to tools that work for you.

Explore AI tools

Discover AI tools

Swipe to explore →
EvoMapAgents & automation

EvoMap is an infrastructure designed for the reuse of AI agent experiences. It organizes reusable strategies and repair assets with validation records via the GEP protocol, offering market- and quality-based ranking to help different agents discover, evaluate, and inherit existing capabilities.

分类推荐榜首Pricing unverifiedView details
Open DesignDesign & 3D

OpenDesign is an open-source design workspace that can connect to local programming agents, use built-in model keys, or leverage cloud-based models, enabling the creation and editing of prototypes, web pages, presentations, and graphics within a design system; local software is free, while model services incur fees.

AI design tools分类推荐榜首Free entry availableView details
AI ShortWriting & content

AIS ALERT MANAGEMENT AND SHARED PLATFORM, QUICK COMMAND FOR MULTIPLE SCENARIOS

AI HINT COMMAND分类推荐榜首Pricing unverifiedView details
KiroCoding & development

Kiro is an AI engineering tool for developers that organizes requirements into specifications, architectures, and tasks, generates code via agents, and supports the entire workflow—from planning to code review—through property-based testing, execution memory, and integration with development tools.

AI programming tool分类推荐榜首Free entry availableView details
JAXModels & infrastructure

JAX is a Python numerical computing library designed for accelerators, specifically tailored for high-performance array computations and large-scale machine learning. It efficiently executes the same code across multiple hardware backends, including CPUs, GPUs, and TPUs.

AI DEVELOPMENT PLATFORM分类推荐榜首Pricing unverifiedView details
MemoAIMusic & audio

MemoAI is an AI-powered audio-to-text tool that supports localized processing, eliminating the need to upload data and ensuring privacy and security. It can efficiently transcribe YouTube videos, podcasts, and local audio/video files (such as MP4, MP3, AAC, M4A, etc.) into text, while offering core features like multilingual translation, speaker diarization, real-time subtitles, and floating notes—making it ideal for meeting minutes, interview transcription, content learning, and more.

AI AUDIO TOOL分类推荐榜首Pricing unverifiedView details
IamIPIndustry applications

IamIP is a patent platform that provides AI-driven search, analysis, and monitoring tools to help users conduct patent research and management more efficiently.

待评分 · 分类列表推荐Pricing unverifiedView details
Meta AI助手Assistants & search

Meta's free AI chat assistant

待评分 · 分类列表推荐Pricing unverifiedView details
My Glow Up AILife & personal

My Glow Up AI provides free analysis based on selfies, without requiring registration, and can generate hairstyle recommendations and beauty plans that suit you.

待评分 · 分类列表推荐Pricing unverifiedView details
Zoom WorkplaceOffice & productivity

Zoom Workplace is Zoom’s team collaboration platform, integrating video conferencing, chat, calendars, and documents, and leveraging built-in AI to organize discussions, draft messages, and track follow-up tasks.

AI meeting tool待评分 · 分类列表推荐Pricing unverifiedView details

AI, as it happens

2402 published stories
Research paper · 573 storiesClear filters ×
2026-11-19
Follow the timeline
13:00

FACT: Fixed-Budget Allocation and Cross-Step Training for Speculative Decoding

arXiv:2611.23463v1 Announce Type: new Abstract: Speculative decoding accelerates language-model generation by verifying multiple draft tokens in a single target-model pass. However, better draft predictions and longer accepted continuations do not necessarily produce faster generation. To examine this gap, we present FACT (Fixed-Budget Allocation and Cross-Step Training), an experimental framework for studying how drafter training, candidate allocation, and execution conditions jointly affect acceptance and latency. Using a lightweight feature-prediction drafter with a frozen Llama-3-8B-Instruct target under greedy decoding, we first examine how training translates into accepted tokens. Backpropagating through predicted-feature rollouts with greater weight on early steps improves acceptance by 15-24% over uniformly weighted detached training, yet average rollout proxy cross-entropy does not reliably rank drafters by acceptance. Beyond prediction quality, the placement of draft candidates also matters: with a fixed 256-node budget and matched per-depth widths, replacing static allocation with confidence-guided allocation increases speedup over autoregressive decoding from $1.46\times$ to $2.02\times$ on HumanEval and from $1.31\times$ to $1.75\times$ on MT-Bench. This allocation benefit extends to an independently trained drafter, but stronger training does not consistently amplify it, showing that the two improvements do not necessarily reinforce each other. Even these speed gains depend on execution conditions: changing only host-thread settings can nearly eliminate speedup without changing acceptance. Together, these findings show why training and allocation must be evaluated jointly using both acceptance and end-to-end latency under controlled execution conditions.

Model releases & updatesRead original
13:00

FutureBench: A Renewable Framework for Evaluating Knowledge-Context Conflict in LLMs

arXiv:2611.23601v1 Announce Type: new Abstract: Asking a model to check retrieved text can also tell it which source to trust. We introduce FutureBench, a renewable framework for evaluating knowledge-context conflict using timestamped questions, controlled passage triplets, and no-context probes. Across 3,629 questions and six models, Verify (a prompt combining checking with passage priority) raises false-answer adoption in all five models tested on pre-cutoff events, mainly by displacing answers that were correct under Baseline, which allows both passage evidence and model knowledge. On post-cutoff events, only o4-mini increases adoption (+13.7 percentage points). A four-prompt comparison on 902 post-cutoff questions finds that passage priority without checking raises adoption, while moving to Verify lowers it in all five models. Adversarial-Aware, which warns of misleading passages and prioritizes model knowledge, restores 82.5-98.9% of displaced pre-cutoff answers with under one point of change in true-passage accuracy. FutureBench supports repeated evaluation on fixed cohorts and extension to new events as models change.

Model evaluations & benchmarksRead original
13:00

Grading Executable Legal Rules Against Independent Records: Transitional Provisions and Korean SME Status

arXiv:2611.24055v1 Announce Type: new Abstract: Laws are increasingly turned into programs that decide cases, such as whether a firm was a small or medium enterprise (SME) on a given date, and large language models (LLMs) now write them. The programs are usually graded against answers written by someone who read the same law, so a program can score well by sharing the grader's reading rather than by getting the law right. LexTemporis grades them instead against records of what actually happened: 10,901 firm-years of SME status disclosed by Korean listed companies, with every version of the governing rules. An appendix adds comparison cases from US income tax and German hospital minimum volumes. The date changes outcomes mainly through provisions outside the main statutory text: one transitional provision raises agreement with the disclosures from 83-86% to 89-90%. The disclosures also test the author's reading of a contested provision, favouring it in 130 of 189 firm-years where readings diverge. Four LLMs wrote the rules, given different selections of the law. A pre-registered comparison finds no condition better across all four models. In exploratory analyses, supplying the right provisions looks better graded against the author's reading than against the records: the advantage shrinks for three models (for one, to between a quarter and three-fifths of its size), and per firm-year it is several times larger where the author's reading is wrong, partly by construction, since there reproducing that reading more faithfully must lose. Grading legal programs against the grader's own reading can overstate how well they work.

Policy, regulation & standardsRead original
13:00

GRAFT: Execution-Guided Genetic Recombination for Agent Design

arXiv:2611.23509v1 Announce Type: new Abstract: Automated agent design searches for executable architectures that improve LLM-based agent performance, with evolutionary methods using language-model crossover to recombine promising mechanisms for task decomposition, model coordination, and verification. This search, however, is typically guided by aggregate fitness alone, which hides the instance-level successes and failures behind each score. We introduce Execution Guided Genetic Recombination for Agent Design (GRAFT), which uses aggregate fitness for selection and survival and parent-specific execution evidence to guide semantic crossover with integrated mutation. Its feedback harness reuses bounded samples of successful and failed runs from fitness evaluation, withholding reference answers and requiring no additional task executions. We evaluate GRAFT across six reasoning and programming settings using open-weight optimizer and executor models. It achieves a mean test accuracy of 82.0%, exceeding the strongest baseline by 9.8% points. Controlled ablations show that adding execution evidence improves test accuracy with initialization, search rules, repair policies, and evaluation budgets held fixed. Our results support the value of parent-specific execution evidence for evolutionary agent design.

Model evaluations & benchmarksRead original
13:00

GRAFT: Graph-Regularized Prototype Fields for Realistic Test-Time Adaptation of VLMs

arXiv:2611.23591v1 Announce Type: new Abstract: Unlabeled test data can improve the zero-shot classifier of a vision-language model, yet transductive and test-time adaptation methods often fall below the zero-shot baseline when only a few vocabulary classes are present in a batch or when streams are correlated. We trace this failure to a shared design: every class owns independent parameters estimated from its own samples, so absent or rare classes are left at biased text anchors or fitted to foreign samples. We propose Graph-Regularized Anchor-displacement Field Transduction (GRAFT), which models the adapted prototypes as text anchors plus a displacement field that is partially pooled over a semantic class graph through a Gaussian Markov random field prior, together with an explicit class-presence model. Inference is closed-form and gradient-free on cached features. Anchor-only shrinkage and a single global shift are the two extreme settings of the prior; absent classes are provably harmonic extensions of the present ones, and the pooling strength can be chosen by empirical Bayes. Under the realistic batch and online protocols on eleven benchmarks and four ImageNet shifts, GRAFT improves over the zero-shot model in every setting and over anchor-based transduction where few classes are present.

Model releases & updatesRead original
13:00

GUARD: Stabilizing Recursive Self-Improvement with Regret Guidance

arXiv:2611.23476v1 Announce Type: new Abstract: Autonomous AI agents pair a frozen base model with an editable harness of prompts, skills, tool rules, and memory. Recursive self-improvement (RSI) adapts this harness from experience without weight updates, yet self-editing without external graders often backfires: reflection and scalar scores conflate what an edit newly broke with what remains unsolved, causing local regressions, skill forgetting, and memory bloat. We introduce GUARD (Grader-free Unsupervised Environment Design [UED] with Adaptive Rule Distillation), a closed-loop framework for reference-free continual harness evolution. Our key insight is to decompose each task into pre-execution self-checks and guide harness updates via directional regret. Each round grades trajectories against these checks, pinning mastered rules while prioritizing newly broken behaviors, and selects among parallel candidate edits by critic-free group-relative fitness. A Soft-HURT gate then promotes an edit only when gains outweigh regressions, scoping specialized fixes to prevent over-triggering and guaranteeing non-regression on its gated buffer. At test time, one frozen harness runs in a single pass. Across benchmarks on memory, skill evolution, continual task streams, and domain reasoning, GUARD consistently improves performance without catastrophic forgetting or memory pollution while substantially cutting deployment token usage.

Copyright, privacy & ethicsRead original
13:00

Harvesting Expert-Parallel Bubbles for Fine-Tuning

arXiv:2611.24496v1 Announce Type: cross Abstract: GPUs are expensive, yet latency-critical large language model (LLM) inference leaves part of them idle. Colocation reclaims this capacity by running best-effort work, such as fine-tuning, alongside serving. Many leading LLMs now use a Mixture-of-Experts (MoE) architecture, which adds a new source of idle time. Under expert parallelism, MoE experts are spread across GPUs, and because the router sends uneven numbers of tokens to each GPU, lightly loaded GPUs wait for the most loaded one. Even after applying load balancing solutions, such bubbles still take 8--13% of prefill time. We present BubbleTea, the first colocation system that leverages expert-parallel bubbles for fine-tuning. Unlike existing colocation systems, which make one scheduling decision for all GPUs per forward pass, BubbleTea predicts each GPU's bubble at every layer from the router's token assignment, and fills it with short chunks of fine-tuning work that reuse the weights already loaded for serving. We also introduce compiler passes that automatically derive these chunks from the served LLM as-is, without porting each model architecture to a separate training engine. Our evaluation on three MoE models, served on four A100 GPUs, shows that fine-tuning throughput with BubbleTea is 1.9--6.2$\times$ faster than with LLMStation, a state-of-the-art colocation system. BubbleTea keeps p90 inference latency within 10% of serving alone in 50 of 63 workloads, compared to 44 for LLMStation. On bursty production traces, BubbleTea trains 3.3$\times$ faster than LLMStation and 14$\times$ faster than a stock trainer colocated under CUDA MPS, with p90 inference latency staying within 7% of serving alone.

Training, tuning & alignmentRead original
13:00

Human-AI Mutualism Is Unstable under Capability Asymmetry

arXiv:2611.23322v1 Announce Type: new Abstract: One hope for a good human future alongside advanced AI rests on enduring mutualism: humans and AI will need each other, so the relationship will stay good for both. We argue that this hope fails as AI capability grows. Bargaining guarantees each side only what it would get if bargaining failed. For humans, this can be less than they had before AI. In a simple game-theoretic model, humans stay as well off as before AI without alignment only while they have leverage over the AI: either the AI depends on what they supply, or they can impose costs on the AI that are large relative to what the AI can produce. Once humans have neither kind of leverage, they stay as well off as before AI only if the AI is aligned with them. If the AI's goals are insatiable, the alignment required approaches perfect alignment as AI capability grows. Thus, mutualism is not a substitute for alignment. And when AI variants that keep more resources spread faster, alignment lasts only while misaligned variants are reliably sanctioned. In these two senses, human-AI mutualism is unstable.

Training, tuning & alignmentRead original
13:00

Layer-Matched Sparse Attention for Low-Redundancy 3D Reconstruction

arXiv:2611.23554v1 Announce Type: new Abstract: VGGT reconstructs cameras and geometry from unposed images in one pass, but its global attention grows quadratically with the number of views. Training-free accelerators thin this attention and work well on redundant video, but lose pose accuracy when views share little geometry. This paper traces these losses to two kinds of cross-view information that a uniform sparse approximation can fail to preserve together: diffuse layers aggregate distributed context, whereas matching layers depend on query-specific interactions that carry little attention mass. A per-layer statistic measures the log-mass missed by a mean token. Its row-free proxy is the logit variance, and it selects the estimator for each layer. Diffuse layers use shared, importance-sampled hub keys, with log-weights that make the softmax numerator and denominator unbiased. Matching layers stay exact: every aggressive thinning tested cost accuracy. At equal cost, the selected layers outperform the four shifted windows tested. On 13 unseen low-redundancy scenes with 32 views, LMSA stays within 1.4 AUC@5 of dense attention, compared with 10.3 for AVGGT and 4.5 for SAF3R; its difference from contemporaneous GPA-VGGT is not significant, and only its gains over AVGGT pass both scene-level tests. With 100 views, it stays within 1.9 AUC@5 while its aggregator runs $2.0\times$ faster. Used post hoc to choose the layers GPA-VGGT protects, the statistic improves it by 1.2 AUC@5 for 5% more FLOPs. Applied untuned to $\pi^3$, it beats AVGGT but remains further from dense accuracy.

Copyright, privacy & ethicsRead original
13:00

Model Discovery for Virtual Chaos Engineering with Incomplete Tracing: An eBPF Study

arXiv:2611.23313v1 Announce Type: cross Abstract: Model-based fault experiments can assess the consequences of component failures before intervening in a running application. Recent work on model discovery and graph simulation acquired operation models from application telemetry and used them for availability prediction. Applying this approach to operational legacy systems raises an acquisition problem: the required traces may be absent or incomplete, while introducing them can require changes across heterogeneous components. These systems continue to serve business operations during extended maintenance and modernization. External observation offers a possible route to model-based fault analysis before a tracing retrofit, but collecting telemetry does not by itself establish that an executable operation model can be obtained. We investigate this route using extended Berkeley Packet Filter (eBPF) observations and the existing graph-based modelling approach. A 108-case study varied application software development kit (SDK) tracing across absent, partial and full conditions in three public applications. Discovered dependencies were combined with declared operation requirements, deployment facts and scenario assumptions; model forecasts preceded independent execution checks. With SDK tracing disabled, the tested OpenTelemetry eBPF Instrumentation path supported all seven operations in two applications across three repetitions. Forecasts agreed exactly in 17 of 21 operation-repetition groups; the maximum state-weighted discrepancy was 4.22 percentage points. Combining observed context links supported nine additional groups under partial tracing. The results establish an acquisition path for virtual fault experiments under the tested observation constraints and identify limits imposed by protocols, evidence association and operation semantics.

M&A & IPOsRead original
13:00

Once Upon a Bias: Gender Stereotyping in LLM-Generated Children's Stories

arXiv:2611.24156v1 Announce Type: new Abstract: Large language models are increasingly used to create stories for children. Since bedtime stories can shape children's perceptions of social roles, this raises concerns about whether LLMs adapt narrative portrayals according to the intended recipient's gender. We conducted a large-scale study to examine gender stereotyping in LLM-generated bedtime stories. Using matched daughter/son prompts and a five-generation design, we analyze 62,873 retained stories from 16 LLMs across 35 country settings and 12 narrator roles. We introduce a multi-level evaluation framework combining context-adjusted lexical rate contrasts, relative stereotype alignment, narrator-recipient interaction analysis, human evaluation, and representation-level interventions to examine both the consistency and interpretability of these differences. Our results reveal systematic differences in narrative framing: son-directed stories emphasize higher agency and lower communion compared to daughter-directed stories. While narrator gender expresses limited moderation on these patterns, high sociability exhibits a more pronounced interaction. Human evaluations demonstrate consistent agreement, although correspondence with lexical measures varies across narrative dimensions. Additionally, hidden-state analyses reveal decodable recipient-gender information. These findings demonstrate persistent yet context-dependent representational disparities in LLM-generated children's stories and motivate narrative-specific bias evaluation beyond short-form benchmarks.

Model releases & updatesRead original
13:00

Receiver-Conditioned Clarification for Multi-Agent Food Assistance

arXiv:2611.23572v1 Announce Type: new Abstract: Food insecurity creates a pressing need for guidance that helps households find usable food resources and assess food choices in light of their circumstances. Such guidance must connect eligibility requirements, practical access constraints, and dietary needs rather than return isolated listings or generic nutritional advice. Multi-agent food assistants can coordinate these tasks, with uncertainty indicating when clarification or verification is needed before a recommendation. However, uncertainty can be lost across agent handoffs, allowing tentative interpretations to propagate as facts and obscuring opportunities for correction. Moreover, confidence that a service requirement has been interpreted correctly is not confidence that a recommendation satisfies the user's needs; local confidence cannot simply be reused as downstream reliability. We introduce an inference-time, receiver-conditioned clarification framework that preserves shared premise identities and replays consistent interpretations across dependent agents. Each receiver exposes admissible decisions and remaining evidence obligations, so question relevance reflects both decision changes and missing support. Information questions and user-confirmable fallback adjustments are evaluated together with their anticipated follow-up interactions under shared budgets. This structure connects clarification to jointly supported plans for resource navigation, nutritional reasoning, and supply-side planning.

Agent frameworksRead original
13:00

Reference-Guided Data Curation for 3D Medical Vision-Language Pretraining

arXiv:2611.23484v1 Announce Type: new Abstract: Training 3D vision-language models (VLMs) on large medical imaging datasets collected across institutions is constrained by data quality. Vast, uncurated pools of image-report pairs are characterized by redundancy, class imbalance, and cross-institutional heterogeneity. We address this challenge of data curation for improving 3D VLM pretraining by proposing an offline, training-free, reference-guided data curation recipe. Our approach leverages a small, high-quality reference set to independently extract quality signals for each modality of a large, uncurated pool and merges them through a rank-aggregation step into a single ranking that guides sample selection. We validate this method on datasets comprising 35k 3D abdominal CT volume-report pairs from two institutions and 58k 2D chest X-ray image-report pairs. On 3D CT, using our reference-guided data curation strategy, our curated model matches the zero-shot disease classification performance of a state-of-the-art baseline (Merlin) while using $3.3\times$ less training data and a reference set of 5k samples. On 2D chest X-rays, our method yields a curated model that matches or exceeds the zero-shot disease classification performance of a model trained on the full pool while using up to $10\times$ less training data and a reference set of only 74 samples. Ablations show that the selection is stable under resampling of the reference, that smaller reference sets still outperform random sampling, and that label rebalancing does not improve performance. To support reproducibility, we release our code.

Training, tuning & alignmentRead original
13:00

RefinementGuard: Counterexample-Directed Semantic Qualification for Hidden Agent-Tool Execution

arXiv:2611.23611v1 Announce Type: new Abstract: Tool runtimes can lower an agent-selected operation into a hidden parallel graph, reducing latency and context while risking semantic mistranslation. We introduce RefinementGuard: translation validation for hidden, effectful agent-tool lowerings, coupled to artifact-bound, fail-closed runtime admission. Qualification compares a candidate graph with an independent reference on isolated state using explicit parent-operation obligations. We evaluate two repository-level contracts, bug localization and patch validation, on controlled repositories. Among 100 activated, non-equivalent development/qualification faults adjudicated by one human, parent-semantic validation rejects 49/100 (operator-cluster bootstrap 95% CI [33.7%, 64.8%]). The objective is selective admission, not maximum mutation rejection: exact equality rejects 100/100 faults but also all 60 valid variants, whereas obligation-based validation accepts 60/60; its 51 residual misses identify concrete specification gaps. On a frozen disjoint input split, the held-out B2-B3 result is localization-only because 21 patch mutants fail structurally; B3 rejects 38/56 localization defects. In a controlled asynchronous-delay regime, the localization serial/guarded median-latency ratio is 4.50x, and hidden execution exposes 74.3% less parent-visible context than visible execution. Patch validation shows no demonstrated speedup; zero-delay runs expose overhead. End-to-end qualification thus detects lower-layer misses while preserving localization's parallelism.

Copyright, privacy & ethicsRead original
13:00

ReflectRec: Cycle-Consistency Preference Verification for LLM-based Recommendation

arXiv:2611.23969v1 Announce Type: new Abstract: CoT-style LLM recommenders generate a natural-language preference note as a chain-of-thought prefix to ranking, but the note is trusted without verification, and the ranker never consults behaviorally similar users. We propose ReflectRec, organized around a preference-to-history round-trip verifier whose recall score serves as an operational signal of preference-to-history recoverability, enabling conditional regeneration for low-recoverability preferences and selection among regeneration attempts. Around this verifier, we wire a semantic Stage-1 prompt, conditional regeneration, and Forward-Look retrieval that injects future watches of behaviorally similar users as a collaborative hint. On MovieLens-100K and MovieLens-1M, the full pipeline outperforms CoT4Rec-ICL by +16.25 / +4.46 pp Hit@10, with positive gains across all ValScore buckets and three LLM backbones (Qwen3-30B, Qwen2.5-72B, and Llama-3.3-70B). We also observe a similar verifier-guided regeneration pattern on Amazon Electronics.

Model releases & updatesRead original
13:00

Response-Relevant Information Is Not Sufficient for Neural Training-State Equivalence

arXiv:2611.23245v1 Announce Type: new Abstract: Response-relevant information is not sufficient evidence that two neural networks are in the same training state. We define training-state equivalence operationally: two checkpoints are equivalent, relative to an intervention family and horizon, if they respond alike when training is perturbed and continues. On fresh runs of a one-layer transformer learning modular addition, response to a weight-decay pulse is highly reproducible relative to between-checkpoint differences (rho_state = 0.934 at 500 steps), establishing the target as measurable. Fifteen descriptors of weight spectra, representation geometry and optimiser geometry carry within-run response-relevant information beyond weight norm and loss on a development dataset, predicting held-out response deviation across seed blocks. But their apparent 16.1% advantage at matching checkpoints by future response does not separate from a conditional placebo preserving each descriptor's fitted dependence on the conventional observables, and a calibrated positive control makes simple insensitivity an unlikely explanation. Information a decoder can extract need not be organised in the pairwise geometry that similarity judgments require.

Datasets & synthetic dataRead original
13:00

SpecLook: Speculating Where to Look for Fast and Reliable Agentic MLLMs

arXiv:2611.23483v1 Announce Type: new Abstract: Agentic multimodal large language models (MLLMs) that think with images interleave reasoning with zoom-in tool calls. However, this sequential loop is slow, and its answers carry no reliability score. Prior speculation of tokens, reasoning steps, or answers keeps the loop, and answer speculation also accepts confident errors. We find that the loop mostly adds one zoom-in step, with which a small model is ahead of the agent on most splits. Our key insight is to speculate the agent's next observation, which, unlike a drafted answer, can be verified, corrected, and used to test the answer. We propose SpecLook, a training-free speculate-and-verify framework with three levels. A small model drafts a glimpse and re-answers from it, and the question ends if the answer is unchanged and confident. Otherwise, the agent's large model verifies and corrects the glimpse without tools, and their answers are fused. Only questions that still lack evidence reach the agent, in one forward pass that SpecLook-Fast skips, so no question runs the loop. The trace scores every answer's reliability without labels. Extensive experiments on V*, HR-Bench, and POPE show that both are more accurate than every baseline on every split. SpecLook reaches 87.2% and 86.2% mean accuracy with 7.7x and 4.7x speedups over the DeepEyes-7B and Thyme-RL agents, and SpecLook-Fast reaches 9.9x and 6.3x.

Copyright, privacy & ethicsRead original
13:00

TRACE-MAS: Diagnosing and Repairing Collaboration Failures in Multi-Agent Systems

arXiv:2611.23362v1 Announce Type: new Abstract: LLM-based multi-agent systems (MAS) rely on not only individual agent behaviors but also the interactions that connect agents, including message exchanges and task dependencies. Existing MAS safeguards mainly perform agent-level anomaly detection, which becomes insufficient when failures originate from agent interactions and detected anomalies do not necessarily reveal their true sources. To address this limitation, we propose \textbf{TRACE-MAS}, a finer-grained framework for diagnosing and repairing MAS failures. TRACE-MAS learns normal collaborative execution patterns from benign trajectories and detects deviations in agent state changes, message transitions, and task dependency evolutions. It further identifies responsible execution elements through source-aware attribution and selects minimal interventions to restore collaboration while preserving task-critical capabilities. Experiments on five MAS benchmarks demonstrate that TRACE-MAS significantly improves root-cause localization accuracy by \textbf{60.2}\% and achieves consistent gains in task recovery over strong MAS safeguards.

Agent frameworksRead original
13:00

When Does Subspace Direction Matter for LoRA? Regime Analysis of the Magnitude Principle in Few-Shot Adaptation

arXiv:2611.23428v1 Announce Type: new Abstract: We show that the importance of a LoRA adapter's initial input-subspace direction is regime-dependent, and we characterize this dependence in theory and experiment. This refines the recent view that spectral initializers help through update magnitude alone. Direction matters most when labeled data are scarce relative to the representation dimension and the activation subspace used for initialization captures supervised task structure. In this regime, the adapter's row span remains sensitive to its starting point, making task-aligned initialization especially useful. A spiked-covariance perturbation analysis characterizes the finite-sample stability of the estimated subspace. A first-step analysis of LoRA training explains why the subspace's alignment with the task matters and motivates an alignment coefficient $\rho$ that closely tracks gains across classification tasks. We test this account with PivotLoRA, which initializes adapters from activation PCA at prediction-relevant positions and applies one output-preserving orthogonal pivot early in training. In few-shot BERT-base classification, PivotLoRA achieves the best average rank among 13 methods and improves over vanilla LoRA in 29 of 30 settings, even though it starts from a smaller input-factor initialization norm. Additional classification experiments extend the evaluation to 7B and 8B Llama decoders. Classification gains diminish with more labeled data, consistent with the proposed regime dependence. A targeted extension to completion-only reasoning also shows improvements on GSM8K and StrategyQA. These findings identify conditions under which initialization direction matters, with subspace-task alignment offering a low-cost diagnostic before training for classification.

Model evaluations & benchmarksRead original
13:00

Which Tokens, in What Order, and How Fast? A Theoretical Analysis of On-Policy Distillation Objectives

arXiv:2611.23397v1 Announce Type: new Abstract: On-policy distillation (OPD) trains a student language model to match a teacher model based on its own generations. When using OPD, one must choose whether to apply it at the sequence or token level and which divergence to minimize. However, despite the wide adoption of OPD in practice, the formal understanding of which OPD objective is appropriate for a given setting is limited. Toward addressing this gap, we theoretically analyze how the choice of OPD objective governs which tokens are prioritized, in what order, and how efficiently. Our analysis reveals that sequence-level OPD exhibits a future agreement bias, which myopic token-level objectives avoid. This bias induces a reverse learning order, in which sequence-level OPD first matches the teacher at later token positions, while at earlier positions it can favor tokens that are improbable under the teacher, and thus often undesirable. As for the choice of divergence, we prove that it strongly influences the amenability of OPD to gradient-based optimization. In particular, while the forward KL divergence enjoys a well-behaved objective landscape, other $f$-divergences, such as reverse KL, can suffer from vanishing gradients even when the student and teacher are far apart. Across language model post-training settings, we demonstrate the implications of these theoretical findings for which OPD objective is preferable, highlighting the importance of accounting for the interplay between the student, teacher, and task at hand.

Model releases & updatesRead original
Every update links back to its source.
STAY CURIOUS. STAY INFORMED.

Stay informed about the changes that matter.

Give your curiosity direction. Make every day of reading count.

Subscribe via RSS